Limits
What Will Not Go Into Words
Some knowledge is a trained judgement rather than a fact, and no amount of incentive, time or skill turns it into a page.
What nobody wrote down here, and what it cost
Judgement about when a client is about to leave
Nobody could write it and nobody tried
Three occasions, all of them noticed too late
Why the architecture is the shape it is
Attempted twice and abandoned
Both drafts described the result rather than the reasons
two things nobody wrote down · 5 occasions it cost somebody something
Two things, five occasions, and no mechanism proposed for either.
A useful operational baseline is to compare the writing burden with real work patterns; this official guide shows one practical way to make that comparison without relying on memory.
A useful independent point of reference for this part of the process is Figma.
Four things one team tried to write down and could not
Knowing when a client is about to leave
Nobody could write it; two attempts were abandoned
Both drafts turned into a list of signs that were individually meaningless
Why the architecture is the shape it is
Attempted twice over two years
Both attempts described the result and stopped where the reasons began
How to judge whether an estimate is realistic
Never attempted
The person who does it well describes it as just knowing
When to stop debugging and roll back
Written as a checklist that nobody follows
The checklist is accurate and does not capture the judgement
four things nobody wrote down · 11 occasions it cost somebody something
Four things, eleven occasions. Every one was attempted or considered by somebody competent and willing. The failures are not failures of effort or of writing ability; they are things whose content is a trained judgement rather than a fact, and this collection has nothing to offer about them. One team's own two years.
The argument so far has been that the obstacle to writing is cost rather than impossibility.
For a small category that is false. Some things cannot be written down by anybody, however willing, and it is worth marking the boundary clearly so that the rest of the argument is not overstated.
What the category contains
Judgement built from many instances. Knowing that a client is going quiet in a way that matters, as opposed to being busy. Knowing that an estimate is optimistic. Knowing when to stop looking and roll back.
In each case the person has seen a hundred examples and has formed something that is not a rule. Asked to explain, they produce a list of signs, each of which is individually useless and some of which are usually absent.
Why the list of signs does not work
Because the knowledge is in the weighting, not in the items.
A written version says: watch for slower replies, shorter messages, fewer people on calls. All true, all present constantly in clients who are perfectly happy, and the skill is in how much each one counts given everything else.
Somebody following the list produces false positives at a rate that makes it worthless. The author, who knows this, either does not write it or writes it and watches it go unused.
The architecture case
Worth separating because it looks like the same thing and is not quite.
"Why the architecture is the shape it is" is usually not tacit. It is lost — the reasons existed, were held by specific people, and were never recorded. The attempts fail because the material is gone, not because it is inexpressible.
That is a different and more tractable problem: it can be prevented in future even where it cannot be recovered now. The remedy is recording reasons at the moment of deciding, which is a cost question like everything else in this collection.
Confusing the two leads teams to treat recoverable losses as unavoidable.
What can be written about the unwritable
Not the judgement, but three things around it.
That it exists, and who holds it. A line saying "judging when a client is at risk is something X does and nobody else currently can" is a document, and it is the most useful one available here.
The cases. Not the rule but the instances: what happened, what the signs were, what was decided and whether it was right. A file of cases is how this kind of knowledge is actually transmitted, slowly, and it can be accumulated by anybody.
The decision process rather than the decision. Who is consulted, what gets looked at, when the judgement is made. This is procedural and writable even when the judgement is not.
How it actually transfers
By apprenticeship, which is slow and is the only mechanism.
Somebody watches the person make the judgement, repeatedly, with the reasoning spoken aloud. Over months they begin to form their own version. This cannot be compressed and cannot be done in writing, and the only thing an organisation can do about it is make the watching possible.
Teams that recognise this arrange for it. Teams that do not keep asking for a document and receiving checklists.
Being honest about the size of this category
It is small, and it is invoked far more often than it applies.
"You can't really write that down" is a sentence used about a great deal of material that is simply inconvenient to write, and treating every difficult explanation as tacit knowledge is a way of not doing the work.
The test: can somebody competent follow a written version and get a reasonable result most of the time? If yes — even with exceptions — it is writable and the difficulty is cost. If the written version produces wrong answers in the hands of a careful reader, it belongs here.
What can be written about the unwritable, at a glance
The judgement does not go into words. Three things around it do, and all three are worth having.
- That it exists, and who holds it. "Judging when a client is at risk is something Marcus does and nobody else currently can" is a document, and the most useful one available here.
- The cases. Not the rule but the instances: what happened, what the signs were, what was decided, whether it was right.
- The process around it. Who is consulted, what gets looked at, when the judgement is made. Procedural and writable even when the judgement is not.
The second is how this kind of knowledge actually transfers, slowly, and anybody can accumulate it.
What this section is for
The remaining pages cover the rest of the residue: things that should not be written, people who will not write them, and the point at which continuing to ask is the wrong move.
None of it is large. All of it is real, and a collection that argued everything was solvable by cheaper writing would be easy to dismiss on the strength of these few cases.